KDE Gear projects with failing CI (master) (21 May 2024)

[Volker Krause]

On Samstag, 25. Mai 2024 13:34:53 CEST Ben Cooksley wrote:
> On Sat, May 25, 2024 at 3:09 AM Volker Krause <vkrause at kde.org> wrote:
> > On Freitag, 24. Mai 2024 12:24:16 CEST Ben Cooksley wrote:
> > > On Fri, May 24, 2024 at 4:13 AM Volker Krause <vkrause at kde.org> wrote:
> > > > On Mittwoch, 22. Mai 2024 10:27:07 CEST Ben Cooksley wrote:
> > > > > On Wed, May 22, 2024 at 9:24 AM Albert Astals Cid <aacid at kde.org>
> > 
> > wrote:
> > > > > > Please work on fixing them, otherwise i will remove the failing CI
> > > > 
> > > > jobs on
> > > > 
> > > > > > their 4th failing week, it is very important that CI is passing
> > > > > > for
> > > > > > multiple
> > > > > > reasons.
> > > > > > 
> > > > > > Good news: 8 repositories fixed
> > > > > > 
> > > > > > Bad news: 6 new repositories started failing
> > > > > > 
> > > > > > 
> > > > > > kio-gdrive - NEW
> > > > > > 
> > > > > >  * https://invent.kde.org/network/kio-gdrive/-/pipelines/693840
> > > > > >  
> > > > > >   * Qt 5.15 is unbuildable because needs libkgapi and we don't
> > 
> > have a
> > 
> > > > Qt5
> > > > 
> > > > > > CI
> > > > > > build of libkgapi anymore. This is REAL BAD.
> > > > > 
> > > > > Looks like kio-gdrive needs to drop the support it has for Qt 5?
> > > > > Can't really see another path forward as libkgapi has no support for
> > 
> > Qt
> > 
> > > > > 5
> > > > > anymore.
> > > > > 
> > > > > This is alas one of those very difficult to solve issues, especially
> > > > > when
> > > > > semi-leaf projects like libkgapi are used more widely.
> > > > 
> > > > Would it work to have a kf5 branch rule for libkgapi pointing to the
> > 
> > last
> > 
> > > > branch with Qt 5 support (and similar for all its dependencies)?
> > > 
> > > Unfortunately not possible i'm afraid - as release/23.08 is no longer
> > > supported (as no further releases are being made).
> > > I therefore terminated all CI support for that branch to ensure that any
> > > issues like this one were forced to the surface.
> > 
> > But do we actually need CI for libkgapi for this? We only need it
> > available as
> > a dependency, so theoretically even distro packages in the CI image would
> > work
> > (I'm very reluctant to try that though, as that might have all kinds of
> > surprising side-effects due to whatever else that might pull in (e.g.
> > ECM)).
> > Therefore the idea to let the seed job provide it, by means of a kf5
> > branch
> > rule.
> 
> Distribution packages definitely won't work :)
> 
> Unfortunately the seed jobs can't help here, as the entirety of
> release/23.08 has been dropped from the CI system.
> There are also rules in place to continually re-drop any release/23.08
> packages that do appear in the system.

ah, so the problem isn't that the seed job wouldn't be able to build this, but 
the result wont be persisted?

Next idea: add a kf5 branch to libkgapi, pointing to the latest release/23.08 
state, and change the branch rules accordingly. We did that for a few non-
branched repos during the transition period of their (branched) consumers IIRC 
(e.g. kweathercore, kirigami-addons).
 
> > > The dependency chain is also @same as both are part of KDE Gear so from
> > > a
> > > technical perspective that doesn't work either.
> > 
> > It's @stable and @latest-kf6 depending on the Qt version in kio-gdrive,
> > and
> > inside libkgapi it's @latest for its KF5 dependencies, which seems correct
> > IIUC?
> 
> @stable for the relationship between libkgapi and kio-gdrive isn't correct
> as they're both within KDE Gear no?

In general, yes. But practically libkgapi isn't in Gear anymore from a Qt5 
kio-gdrive perspective, so pointing to a specific (older) branch instead would 
seem like something that could help here.

> > > From a practical perspective, i'm not sure you can really release
> > 
> > something
> > 
> > > as part of a bundle that needs something from an older release of that
> > 
> > same
> > 
> > > bundle in order to build...
> > 
> > We already have that mixed scenario anyway in 24.02 it seems, so that is
> > apparently working.
> 
> I'm only aware of one case where that happened, which was a special one off
> as KF5 apps needed the Qt 5 plugins still?
> (I think this was kio-extras)

Yeah, KIO workers is one such case. There's also some inter-dependencies in 
edu that are not a problem yet but where different parts seem to be moving at 
different speeds towards Qt 6 (Labplot <-> Cantor, Marble <-> *). Hopefully 
none of that becomes a problem, but I'd feel better if we have less drastic 
options available to deal with that :)

> Perhaps we need to ask distributions to treat kio-gdrive the same?
> 
> > > The only fix is to either drop Qt 5 support from kio-gdrive, or to
> > 
> > restore
> > 
> > > Qt 5 support to libkgapi.
> > 
> > Don't get me wrong, I'm all for retiring Qt5 stuff as soon as we can, but
> > I
> > fear this isn't the only place where we will hit this problem, and not all
> > of
> > those might be able to do that just yet.
> 
> In an ideal world, applications (the leaves) would drop Qt 5 support first,
> and then once all consumers had migrated away the libraries would start
> dropping support.
> Not sure why libkgapi had to rush to drop Qt 5....

My guess is this was mainly a matter of not being aware of kio-gdrive using 
this and only considering kdepim-runtime for the decision.

Regards,
Volker
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/kde-devel/attachments/20240526/49e6c9df/attachment.sig>